Wieslaw L. Nowinski

Medical Image Segmentation Using K-Means Clustering and Improved Watershed Algorithm

We propose a methodology that incorporates k-means and improved watershed segmentation algorithm for medical image segmentation. The use of the conventional watershed algorithm for medical image analysis is widespread because of its advantages, such as always being able to produce a complete division of the image. However, its drawbacks include over-segmentation and sensitivity to false edges. We address the drawbacks of the conventional watershed algorithm when it is applied to medical images by using k-means clustering to produce a primary segmentation of the image before we apply our improved watershed segmentation algorithm to it. The k-means clustering is an unsupervised learning algorithm, while the improved watershed segmentation algorithm makes use of automated thresholding on the gradient magnitude map and post-segmentation merging on the initial partitions to reduce the number of false edges and over-segmentation. By comparing the number of partitions in the segmentation maps of 50 images, we showed that our proposed methodology produced segmentation maps which have 92% fewer partitions than the segmentation maps produced by the conventional watershed algorithm

Multiple brain atlas database and atlas‐based neuroimaging system

For the purpose of developing multiple, complementary, fully labeled electronic brain atlases and an atlas-based neuroimaging system for analysis, quantification, and real-time manipulation of cerebral structures in two and three dimensions, we have digitized, enhanced, segmented, and labeled the following print brain atlases: Co-Planar Stereotaxic Atlas of the Human Brain by Talairach and Tournoux, Atlas for Stereotaxy of the Human Brain by Schaltenbrand and Wahren, Referentially Oriented Cerebral MRI Anatomy by Talairach and Tournoux, and Atlas of the Cerebral Sulci by Ono, Kubik, and Abernathey. Three-dimensional extensions of these atlases have been developed as well. All two- and three-dimensional atlases are mutually preregistered and may be interactively registered with an actual patients data. An atlas-based neuroimaging system has been developed that provides support for reformatting, registration, visualization, navigation, image processing, and quantification of clinical data. The anatomical index contains about 1,000 structures and over 400 sulcal patterns. Several new applications of the brain atlas database also have been developed, supported by various technologies such as virtual reality, the Internet, and electronic publishing. Fusion of information from multiple atlases assists the user in comprehensively understanding brain structures and identifying and quantifying anatomical regions in clinical data. The multiple brain atlas database and atlas-based neuroimaging system have substantial potential impact in stereotactic neurosurgery and radiotherapy by assisting in visualization and real-time manipulation in three dimensions of anatomical structures, in quantitative neuroradiology by allowing interactive analysis of clinical data, in three-dimensional neuroeducation, and in brain function studies.

On minimum variance thresholding

Variance-based thresholding methods could be biased from the threshold found by expert and the underlying mechanism responsible for this bias is explored in this paper. An analysis on the minimum class variance thresholding (MCVT) and the Otsu method, which minimizes the within-class variance, is carried out. It turns out that the bias for the Otsu method is due to differences in class variances or class probabilities and the resulting threshold is biased towards the component with larger class variance or larger class probability. The MCVT method is found to be similar to the minimum error thresholding.

A rapid algorithm for robust and automatic extraction of the midsagittal plane of the human cerebrum from neuroimages based on local symmetry and outlier removal

A rapid algorithm for robust, accurate, and automatic extraction of the midsagittal plane (MSP) of the human cerebrum from normal and pathological neuroimages is proposed. The MSP is defined as a plane formed from the interhemispheric fissure line segments having the dominant orientation. The algorithm extracts the MSP in four steps: (1) determine suitable axial slices for processing, (2) localize the fissure line segments on them, (3) select inliers from the extracted fissure line segments through histogram-based outlier removal, and (4) calculate the equation of the MSP from the selected inliers. The fissure line segments are localized by minimizing the local symmetry index characterizing anatomical properties of images in the vicinity of the interhemispheric fissure. A two-stage angular and distance outlier removal is introduced to handle abnormalities. The algorithm has been validated quantitatively with 125 structural MRI and CT cases from 10 centers on three continents by studying its accuracy; tolerance to rotation, noise, asymmetry, and bias field; sensitivity to parameters; and performance. A statistical relationship between algorithm accuracy and the datas adherence to planarity is also determined. The algorithm extracts the MSP below 6 s on Pentium 4 (2.4 GHz) with the average angular and distance errors of (0.40 degrees; 0.63 mm) for normal and (0.59 degrees; 0.73 mm) for pathological cases. The robustness to noise, asymmetry, rotation, and bias field is achieved by extracting the MSP based on the dominant orientation and local symmetry index. A low computational cost results from applying simple operations capturing intrinsic anatomic features, constraining the searching space to the local vicinity of the interhemispheric fissure, and formulating a noniterative algorithm with a coarse and fine fixed-step searching. In comparison to the existing methods, our algorithm is much faster, performs accurately and robustly for a wide range of diversified data, and is fully automatic and thoroughly validated, which make it suitable for clinical applications.

Computer-aided stereotactic functional neurosurgery enhanced by the use of the multiple brain atlas database

Introduces a computer-aided atlas-based functional neurosurgery methodology and describes NeuroPlanner, a software system which supports it. NeuroPlanner provides four groups of functions: (1) data-related for data reading, interpolation, reformatting, and image processing; (2) atlas-related for multiple atlases reading, atlas-to-data global and local registrations, two way anatomical indexing, and multiple labeling in two and three dimensions; (3) atlas data exploration-related for three-dimensional (3 D) display and real-time manipulation of cerebral structures, continuous navigation, two-dimensional (2-D), triplanar, 3-D presentations, and 2-D interaction in four views; and (4) neurosurgery-related for targeting, trajectory planning, mensuration, simulating the insertion of microelectrode, and simulating therapeutic lesioning. All operations, excluding atlas and data reading, are real time. The combined anatomical index of the multiple brain atlas database containing complementary 2-D and 3-D atlases has about 1000 structures per hemisphere, and over 400 sulcal patterns. Neurosurgical planning with mutually preregistered multiple brain atlases in all three orthogonal orientations is novel. The approach is validated with 24 intraoperative and postoperative datasets for thalamotomies, thalamic stimulations, pallidotomies, and pallidal stimulations. Its potential benefits include increased accuracy of target definition, reduced time of the surgical procedure by decreasing the number of tracts, facilitated planning of more sophisticated trajectories, lowered cost by reducing the number of microelectrodes used, reduced surgical complications, and the extra degree of confidence given to the neurosurgeon.

Thresholding based on variance and intensity contrast

A new thresholding criterion is formulated for segmenting small objects by exploring the knowledge about intensity contrast. It is the weighted sum of within-class variance and intensity contrast between the object and background. Theoretical bounds of the weight are given for the uniformly distributed background and object, followed by the procedure to estimate the weight from prior knowledge. Tests against two real and two synthetic images show that small objects can be extracted successfully irrespective of the complexity of background and difference in class sizes.

Statistical analysis of 168 bilateral subthalamic nucleus implantations by means of the probabilistic functional atlas.

OBJECTIVE: Statistical analysis of patients previously operated on may improve our methods of performing subsequent surgical procedures. In this article, we introduce a method for studying the functional properties of cerebral structures from electrophysiological and neuroimaging data by using the probabilistic functional atlas (PFA). The PFA provides a spatial distribution of the clinically most effective contacts normalized to a common space. This distribution is converted into a probability function for a given point in space to be inside an effective contact. The PFA was used to analyze spatial properties of the functional subthalamic nucleus (STN), defined as the spatial volume corresponding to the distribution of effective contacts. These results may potentially be useful in planning subthalamic implantation of electrodes. METHODS: In all, 168 bilateral subthalamic stimulations were examined. An algorithm was developed for converting these data into the PFA. The PFA for the STN (here called “atlas”) was calculated with 0.25-mm3 resolution, and several features characterizing the left and right STN were studied. The analysis was performed with and without lateral compensation against the width of the third ventricle. The key feature introduced here used for analysis of the functional STN is a (probabilistic) functional volume of structure (defined for a given probability as the volume of a region whose every point has a probability equal to or greater than this given probability). RESULTS: The analysis has been performed for two situations: with and without lateral compensation against the width of the third ventricle. Without lateral compensation, the differences between the mean values and standard deviations of their barycenter coordinates for the left and right functional STNs are 0.31 and 0.18 mm, respectively. The left STN and right STN exhibit differences in functional volume size and probability distribution. The entire functional volume is 240 mm3 for the left and 229 mm3 for the right STN. A more prominent difference exists in the region of high probabilities (0.7 or higher), called “hot STN.” The volume of the left hot STN is 5.52 mm3, whereas that of the right is 3.92 mm3. The left hot STN is 1.41 times bigger and 20% more dense than the right hot STN. For a given probability, the corresponding functional volume for the left hot STN is up to 43 times larger than that for the right STN. Practically speaking, lateral compensation does not change these results qualitatively. Quantitatively, differentiation between the left and right STNs is lower. For instance, for a given probability, the corresponding functional volume for the left hot STN is only up to 11 times larger than that for the right one. In either situation (i.e., with and without lateral compensation), the size of the hot STN in relation to the whole STN remains very small (1–2%). In addition, statistical analysis shows that in either situation, the means of the left and right functional STNs are significantly different. CONCLUSION: PFA-based planning may be superior to the current practice of using anatomic atlases that provide delineation of the target structure only, because it is more precise and provides a unique target point in the stereotactic space. This best stereotactic target is the point in the individualized atlas with the highest probability, meaning the highest probability of having the best target on the basis of the patients previously operated on. This best target is located in the hot STN, the size of which determines the precision of targeting. Because the size of the hot STN in comparison to the whole STN remains very small (1–2%) independent of whether or not lateral compensation is applied, target planning and execution have to be performed with high precision. The methodology presented, based on the PFA and on the functional volume, is general and can be applied to other structures and data sets. As numerous centers keep gathering large amounts of electrophysiological human and animal data, this work may facilitate opening new avenues in exploiting these data.

Supervised range-constrained thresholding

A novel thresholding approach to confine the intensity frequency range of the object based on supervision is introduced. It consists of three steps. First, the region of interest (ROI) is determined in the image. Then, from the histogram of the ROI, the frequency range in which the proportion of the background to the ROI varies is estimated through supervision. Finally, the threshold is determined by minimizing the classification error within the constrained variable background range. The performance of the approach has been validated against 54 brain MR images, including images with severe intensity inhomogeneity and/or noise, CT chest images, and the Cameraman image. Compared with nonsupervised thresholding methods, the proposed approach is substantially more robust and more reliable. It is also computationally efficient and can be applied to a wide class of computer vision problems, such as character recognition, fingerprint identification, and segmentation of a wide variety of medical images.

Fast connected-component labelling in three-dimensional binary images based on iterative recursion

We propose two new methods to label connected components based on iterative recursion: one directly labels an original binary image while the other labels the boundary voxels followed by one-pass labelling of non-boundary object voxels. The novelty of the proposed methods is a fast labelling of large datasets without stack overflow and a flexible trade-off between speed and memory. For each iterative recursion: (1) the original volume is scanned in the raster order and an initially unlabelled object voxel v is selected, (2) a sub-volume with a user-defined size is formed around the selected voxel v, (3) within this sub-volume all object voxels 26-connected to v are labelled using iterations; and (4) subsequent iterative recursions are initiated from those border object voxels of the sub-volume that are 26-connected to v. Our experiments show that the time-memory trade-off is that the decrease in the execution time by one-third requires the increase in memory size by 3 orders. This trade-off is controlled by the user by changing the size of the sub-volume. Experiments on large three-dimensional brain phantom datasets (362x432x362 voxels of 56 MB (megabytes)) show that the proposed methods are three times faster on the average (with the maximum speedup of 10) than the existing iterative methods based on label equivalences with less than 1 MB memory consumption. Moreover, our algorithms are applicable to any dimensional data and are less dependant on the geometric complexity of connected components.

An Algorithm for Rapid Calculation of a Probabilistic Functional Atlas of Subcortical Structures from Electrophysiological Data Collected during Functional Neurosurgery Procedures

The paper introduces an optimal algorithm for rapid calculation of a probabilistic functional atlas (PFA) of subcortical structures from data collected during functional neurosurgery procedures. The PFA is calculated based on combined intraoperative electrophysiology, pre- and intraoperative neuroimaging, and postoperative neurological verification. The algorithm converts the coordinates of the neurologically most effective contacts into probabilistic functional maps taking into account the geometry of a stimulating electrode. The PFA calculation comprises the reconstruction of the contact coordinates from two orthogonal projections, normalizing (warping) the contacts modeled as cylinders, voxelizing the contact models, calculating the atlas, and computing probability. In addition, an analytical representation of the PFA is formulated based on Gaussian modeling. The initial PFA has been calculated from the data collected during the treatment of 274 Parkinsons disease patients, most of them operated bilaterally (487 operated hemispheres). It contains the most popular stereotactic targets, the subthalamic nucleus, globus pallidus internus, and ventral intermedius nucleus. The key application of the algorithm is targeting in stereotactic and functional neurosurgery, and it also can be employed in human and animal brain research.